摘要
将遗传算法与正交优选法结合 ,用来训练径向基函数 ( RBF)神经网络 ,并对基函数宽度进行自动地调整 ,得到了一种训练 RBF神经网络的新方法 .将其应用于连续流体搅拌反应槽 ( CFSTR)生化反应器的建模中 ,得到了令人满意的结果 .该算法提高了径向基函数神经网络的泛化能力和鲁棒性 ,研究表明是一种有效的“黑箱”
Based on the combination of genetic algorithms and orthogonal optimum seeking method, along with the modification of the width of the radial basis function (RBF) a novel nonlinear dynamic system modeling approach is proposed to correctly determine the centers of the RBF. This approach has greatly improved the generalization and robustness of the RBF neural network. Simulation shows that the improved neural network has successfully modeled the continuous flow stirred tank reactor system.
出处
《系统工程理论与实践》
EI
CSCD
北大核心
2001年第3期64-69,共6页
Systems Engineering-Theory & Practice
关键词
RBF神经网络
非线性动态系统
建模
遗传算法
genetic algorithms
radial basis function neural networks
orthogonal optimum seeking method
continuous flow stirred tank reactors
